41 lines (27 loc) · 1.2 KB

GuardMesh Alpha Stability Policy

GuardMesh is still in alpha. The project aims to be stable enough for experimentation, but not yet frozen.

Stability expectations

Most stable in alpha

core decision names: allow, deny, escalate, kill
file-based pack workflow using policy.yaml and request.json
example-pack verification model using decision.expected.json

Moderately stable in alpha

CLI command names
exported runtime helpers
PolicyEngine shape
current rule selectors and constraints

These may still evolve when clarity or correctness improves.

Least stable in alpha

lint heuristics
composition details beyond basic extends
any future bundle or adapter interfaces
exact report field additions for authoring and tooling features

Compatibility approach

Prefer additive changes where possible.
Document behavior changes in README.md and docs/policy-spec.md.
Update example packs and tests in the same change whenever semantics move.

Practical guidance

If you are embedding GuardMesh during alpha:

pin the package version
rely on documented fields rather than incidental internals
treat minor alpha releases as potentially behavior-changing